If my conhjecture was correct, I think the PX qref latch
is the bottleneck "before" the PX Deq Credit: send blkd -
You may find that when you decrease the degree
you start to get more 'send blkd'.
--
Jonathan Lewis
http://www.jlcomp.demon.co.uk
Host to The Co-Operative Oracle Users' FAQ
http://www.jlcomp.demon.co.uk/faq/ind_faq.html
Author of:
Practical Oracle 8i: Building Efficient Databases
Screen saver or Life saver: http://www.ud.com
Use spare CPU to assist in cancer research.
Herman de Boer wrote in message ...
>Thanks for answering.
>
>while watching 'top' for a while, one can see that the load for each
>PQ slave is about equal. Difficult to check with v$pq_tqstat, because
>its only the query coordinator who can see this data - and that one is
>submitted using a scheduler.
>
>What I would've expected, is much more wait time in events like "PX
>Deq Credit: free buffer", "PX Deq Credit: send blkd". There is some
>amount of wait time in "PX Deq: Table Q Normal", which is normal.
>
>I am suspicious abut the data for this event is being measured or
>calculated wrong (probably OS (hp-ux 64-bit) and/or version specific).
>
>Sad that the documentation for this part is far from complete - or
>wrong.
>
>But, decreasing the degree of parallelism looks like a good
>alternative.
>
>
>Kind Regards,
>
>Herman de Boer
>IT Consultancy Group bv
Received on Fri Nov 02 2001 - 16:40:21 CST
